Skip to content

Neo4j

Neo4jGraphStore #

Bases: GraphStore

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
class Neo4jGraphStore(GraphStore):
    def __init__(
        self,
        username: str,
        password: str,
        url: str,
        database: str = "neo4j",
        node_label: str = "Entity",
        **kwargs: Any,
    ) -> None:
        self.node_label = node_label
        self._driver = neo4j.GraphDatabase.driver(url, auth=(username, password))
        self._database = database
        self.schema = ""
        self.structured_schema: Dict[str, Any] = {}
        # Verify connection
        try:
            with self._driver as driver:
                driver.verify_connectivity()
        except neo4j.exceptions.ServiceUnavailable:
            raise ValueError(
                "Could not connect to Neo4j database. "
                "Please ensure that the url is correct"
            )
        except neo4j.exceptions.AuthError:
            raise ValueError(
                "Could not connect to Neo4j database. "
                "Please ensure that the username and password are correct"
            )
        # Set schema
        try:
            self.refresh_schema()
        except neo4j.exceptions.ClientError:
            raise ValueError(
                "Could not use APOC procedures. "
                "Please ensure the APOC plugin is installed in Neo4j and that "
                "'apoc.meta.data()' is allowed in Neo4j configuration "
            )
        # Create constraint for faster insert and retrieval
        try:  # Using Neo4j 5
            self.query(
                """
                CREATE CONSTRAINT IF NOT EXISTS FOR (n:%s) REQUIRE n.id IS UNIQUE;
                """
                % (self.node_label)
            )
        except Exception:  # Using Neo4j <5
            self.query(
                """
                CREATE CONSTRAINT IF NOT EXISTS ON (n:%s) ASSERT n.id IS UNIQUE;
                """
                % (self.node_label)
            )

    @property
    def client(self) -> Any:
        return self._driver

    def get(self, subj: str) -> List[List[str]]:
        """Get triplets."""
        query = """
            MATCH (n1:%s)-[r]->(n2:%s)
            WHERE n1.id = $subj
            RETURN type(r), n2.id;
        """

        prepared_statement = query % (self.node_label, self.node_label)

        with self._driver.session(database=self._database) as session:
            data = session.run(prepared_statement, {"subj": subj})
            return [record.values() for record in data]

    def get_rel_map(
        self, subjs: Optional[List[str]] = None, depth: int = 2, limit: int = 30
    ) -> Dict[str, List[List[str]]]:
        """Get flat rel map."""
        # The flat means for multi-hop relation path, we could get
        # knowledge like: subj -> rel -> obj -> rel -> obj -> rel -> obj.
        # This type of knowledge is useful for some tasks.
        # +-------------+------------------------------------+
        # | subj        | flattened_rels                     |
        # +-------------+------------------------------------+
        # | "player101" | [95, "player125", 2002, "team204"] |
        # | "player100" | [1997, "team204"]                  |
        # ...
        # +-------------+------------------------------------+

        rel_map: Dict[Any, List[Any]] = {}
        if subjs is None or len(subjs) == 0:
            # unlike simple graph_store, we don't do get_all here
            return rel_map

        query = (
            f"""MATCH p=(n1:{self.node_label})-[*1..{depth}]->() """
            f"""WHERE toLower(n1.id) IN {[subj.lower() for subj in subjs] if subjs else []}"""
            "UNWIND relationships(p) AS rel "
            "WITH n1.id AS subj, p, apoc.coll.flatten(apoc.coll.toSet("
            "collect([type(rel), endNode(rel).id]))) AS flattened_rels "
            f"RETURN subj, collect(flattened_rels) AS flattened_rels LIMIT {limit}"
        )

        data = list(self.query(query, {"subjs": subjs}))
        if not data:
            return rel_map

        for record in data:
            rel_map[record["subj"]] = record["flattened_rels"]
        return rel_map

    def upsert_triplet(self, subj: str, rel: str, obj: str) -> None:
        """Add triplet."""
        query = """
            MERGE (n1:`%s` {id:$subj})
            MERGE (n2:`%s` {id:$obj})
            MERGE (n1)-[:`%s`]->(n2)
        """

        prepared_statement = query % (
            self.node_label,
            self.node_label,
            rel.replace(" ", "_").upper(),
        )

        with self._driver.session(database=self._database) as session:
            session.run(prepared_statement, {"subj": subj, "obj": obj})

    def delete(self, subj: str, rel: str, obj: str) -> None:
        """Delete triplet."""

        def delete_rel(subj: str, obj: str, rel: str) -> None:
            with self._driver.session(database=self._database) as session:
                session.run(
                    (
                        "MATCH (n1:{})-[r:{}]->(n2:{}) WHERE n1.id = $subj AND n2.id"
                        " = $obj DELETE r"
                    ).format(self.node_label, rel, self.node_label),
                    {"subj": subj, "obj": obj},
                )

        def delete_entity(entity: str) -> None:
            with self._driver.session(database=self._database) as session:
                session.run(
                    "MATCH (n:%s) WHERE n.id = $entity DELETE n" % self.node_label,
                    {"entity": entity},
                )

        def check_edges(entity: str) -> bool:
            with self._driver.session(database=self._database) as session:
                is_exists_result = session.run(
                    "MATCH (n1:%s)--() WHERE n1.id = $entity RETURN count(*)"
                    % (self.node_label),
                    {"entity": entity},
                )
                return bool(list(is_exists_result))

        delete_rel(subj, obj, rel)
        if not check_edges(subj):
            delete_entity(subj)
        if not check_edges(obj):
            delete_entity(obj)

    def refresh_schema(self) -> None:
        """
        Refreshes the Neo4j graph schema information.
        """
        node_properties = [el["output"] for el in self.query(node_properties_query)]
        rel_properties = [el["output"] for el in self.query(rel_properties_query)]
        relationships = [el["output"] for el in self.query(rel_query)]

        self.structured_schema = {
            "node_props": {el["labels"]: el["properties"] for el in node_properties},
            "rel_props": {el["type"]: el["properties"] for el in rel_properties},
            "relationships": relationships,
        }

        # Format node properties
        formatted_node_props = []
        for el in node_properties:
            props_str = ", ".join(
                [f"{prop['property']}: {prop['type']}" for prop in el["properties"]]
            )
            formatted_node_props.append(f"{el['labels']} {{{props_str}}}")

        # Format relationship properties
        formatted_rel_props = []
        for el in rel_properties:
            props_str = ", ".join(
                [f"{prop['property']}: {prop['type']}" for prop in el["properties"]]
            )
            formatted_rel_props.append(f"{el['type']} {{{props_str}}}")

        # Format relationships
        formatted_rels = [
            f"(:{el['start']})-[:{el['type']}]->(:{el['end']})" for el in relationships
        ]

        self.schema = "\n".join(
            [
                "Node properties are the following:",
                ",".join(formatted_node_props),
                "Relationship properties are the following:",
                ",".join(formatted_rel_props),
                "The relationships are the following:",
                ",".join(formatted_rels),
            ]
        )

    def get_schema(self, refresh: bool = False) -> str:
        """Get the schema of the Neo4jGraph store."""
        if self.schema and not refresh:
            return self.schema
        self.refresh_schema()
        logger.debug(f"get_schema() schema:\n{self.schema}")
        return self.schema

    def query(self, query: str, param_map: Optional[Dict[str, Any]] = {}) -> Any:
        with self._driver.session(database=self._database) as session:
            result = session.run(query, param_map)
            return [d.data() for d in result]

get #

get(subj: str) -> List[List[str]]

Get triplets.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
def get(self, subj: str) -> List[List[str]]:
    """Get triplets."""
    query = """
        MATCH (n1:%s)-[r]->(n2:%s)
        WHERE n1.id = $subj
        RETURN type(r), n2.id;
    """

    prepared_statement = query % (self.node_label, self.node_label)

    with self._driver.session(database=self._database) as session:
        data = session.run(prepared_statement, {"subj": subj})
        return [record.values() for record in data]

get_rel_map #

get_rel_map(subjs: Optional[List[str]] = None, depth: int = 2, limit: int = 30) -> Dict[str, List[List[str]]]

Get flat rel map.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
def get_rel_map(
    self, subjs: Optional[List[str]] = None, depth: int = 2, limit: int = 30
) -> Dict[str, List[List[str]]]:
    """Get flat rel map."""
    # The flat means for multi-hop relation path, we could get
    # knowledge like: subj -> rel -> obj -> rel -> obj -> rel -> obj.
    # This type of knowledge is useful for some tasks.
    # +-------------+------------------------------------+
    # | subj        | flattened_rels                     |
    # +-------------+------------------------------------+
    # | "player101" | [95, "player125", 2002, "team204"] |
    # | "player100" | [1997, "team204"]                  |
    # ...
    # +-------------+------------------------------------+

    rel_map: Dict[Any, List[Any]] = {}
    if subjs is None or len(subjs) == 0:
        # unlike simple graph_store, we don't do get_all here
        return rel_map

    query = (
        f"""MATCH p=(n1:{self.node_label})-[*1..{depth}]->() """
        f"""WHERE toLower(n1.id) IN {[subj.lower() for subj in subjs] if subjs else []}"""
        "UNWIND relationships(p) AS rel "
        "WITH n1.id AS subj, p, apoc.coll.flatten(apoc.coll.toSet("
        "collect([type(rel), endNode(rel).id]))) AS flattened_rels "
        f"RETURN subj, collect(flattened_rels) AS flattened_rels LIMIT {limit}"
    )

    data = list(self.query(query, {"subjs": subjs}))
    if not data:
        return rel_map

    for record in data:
        rel_map[record["subj"]] = record["flattened_rels"]
    return rel_map

upsert_triplet #

upsert_triplet(subj: str, rel: str, obj: str) -> None

Add triplet.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
def upsert_triplet(self, subj: str, rel: str, obj: str) -> None:
    """Add triplet."""
    query = """
        MERGE (n1:`%s` {id:$subj})
        MERGE (n2:`%s` {id:$obj})
        MERGE (n1)-[:`%s`]->(n2)
    """

    prepared_statement = query % (
        self.node_label,
        self.node_label,
        rel.replace(" ", "_").upper(),
    )

    with self._driver.session(database=self._database) as session:
        session.run(prepared_statement, {"subj": subj, "obj": obj})

delete #

delete(subj: str, rel: str, obj: str) -> None

Delete triplet.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
def delete(self, subj: str, rel: str, obj: str) -> None:
    """Delete triplet."""

    def delete_rel(subj: str, obj: str, rel: str) -> None:
        with self._driver.session(database=self._database) as session:
            session.run(
                (
                    "MATCH (n1:{})-[r:{}]->(n2:{}) WHERE n1.id = $subj AND n2.id"
                    " = $obj DELETE r"
                ).format(self.node_label, rel, self.node_label),
                {"subj": subj, "obj": obj},
            )

    def delete_entity(entity: str) -> None:
        with self._driver.session(database=self._database) as session:
            session.run(
                "MATCH (n:%s) WHERE n.id = $entity DELETE n" % self.node_label,
                {"entity": entity},
            )

    def check_edges(entity: str) -> bool:
        with self._driver.session(database=self._database) as session:
            is_exists_result = session.run(
                "MATCH (n1:%s)--() WHERE n1.id = $entity RETURN count(*)"
                % (self.node_label),
                {"entity": entity},
            )
            return bool(list(is_exists_result))

    delete_rel(subj, obj, rel)
    if not check_edges(subj):
        delete_entity(subj)
    if not check_edges(obj):
        delete_entity(obj)

refresh_schema #

refresh_schema() -> None

Refreshes the Neo4j graph schema information.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
def refresh_schema(self) -> None:
    """
    Refreshes the Neo4j graph schema information.
    """
    node_properties = [el["output"] for el in self.query(node_properties_query)]
    rel_properties = [el["output"] for el in self.query(rel_properties_query)]
    relationships = [el["output"] for el in self.query(rel_query)]

    self.structured_schema = {
        "node_props": {el["labels"]: el["properties"] for el in node_properties},
        "rel_props": {el["type"]: el["properties"] for el in rel_properties},
        "relationships": relationships,
    }

    # Format node properties
    formatted_node_props = []
    for el in node_properties:
        props_str = ", ".join(
            [f"{prop['property']}: {prop['type']}" for prop in el["properties"]]
        )
        formatted_node_props.append(f"{el['labels']} {{{props_str}}}")

    # Format relationship properties
    formatted_rel_props = []
    for el in rel_properties:
        props_str = ", ".join(
            [f"{prop['property']}: {prop['type']}" for prop in el["properties"]]
        )
        formatted_rel_props.append(f"{el['type']} {{{props_str}}}")

    # Format relationships
    formatted_rels = [
        f"(:{el['start']})-[:{el['type']}]->(:{el['end']})" for el in relationships
    ]

    self.schema = "\n".join(
        [
            "Node properties are the following:",
            ",".join(formatted_node_props),
            "Relationship properties are the following:",
            ",".join(formatted_rel_props),
            "The relationships are the following:",
            ",".join(formatted_rels),
        ]
    )

get_schema #

get_schema(refresh: bool = False) -> str

Get the schema of the Neo4jGraph store.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/base.py
244
245
246
247
248
249
250
def get_schema(self, refresh: bool = False) -> str:
    """Get the schema of the Neo4jGraph store."""
    if self.schema and not refresh:
        return self.schema
    self.refresh_schema()
    logger.debug(f"get_schema() schema:\n{self.schema}")
    return self.schema

Neo4jPropertyGraphStore #

Bases: PropertyGraphStore

Neo4j Property Graph Store.

This class implements a Neo4j property graph store.

If you are using local Neo4j instead of aura, here's a helpful command for launching the docker container:

docker run \
    -p 7474:7474 -p 7687:7687 \
    -v $PWD/data:/data -v $PWD/plugins:/plugins \
    --name neo4j-apoc \
    -e NEO4J_apoc_export_file_enabled=true \
    -e NEO4J_apoc_import_file_enabled=true \
    -e NEO4J_apoc_import_file_use__neo4j__config=true \
    -e NEO4JLABS_PLUGINS=\\[\"apoc\"\\] \
    neo4j:latest

Parameters:

Name Type Description Default
username str

The username for the Neo4j database.

required
password str

The password for the Neo4j database.

required
url str

The URL for the Neo4j database.

required
database Optional[str]

The name of the database to connect to. Defaults to "neo4j".

'neo4j'

Examples:

pip install llama-index-graph-stores-neo4j

from llama_index.core.indices.property_graph import PropertyGraphIndex
from llama_index.graph_stores.neo4j import Neo4jPropertyGraphStore

# Create a Neo4jPropertyGraphStore instance
graph_store = Neo4jPropertyGraphStore(
    username="neo4j",
    password="neo4j",
    url="bolt://localhost:7687",
    database="neo4j"
)

# create the index
index = PropertyGraphIndex.from_documents(
    documents,
    property_graph_store=graph_store,
)
Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
class Neo4jPropertyGraphStore(PropertyGraphStore):
    r"""
    Neo4j Property Graph Store.

    This class implements a Neo4j property graph store.

    If you are using local Neo4j instead of aura, here's a helpful
    command for launching the docker container:

    ```bash
    docker run \
        -p 7474:7474 -p 7687:7687 \
        -v $PWD/data:/data -v $PWD/plugins:/plugins \
        --name neo4j-apoc \
        -e NEO4J_apoc_export_file_enabled=true \
        -e NEO4J_apoc_import_file_enabled=true \
        -e NEO4J_apoc_import_file_use__neo4j__config=true \
        -e NEO4JLABS_PLUGINS=\\[\"apoc\"\\] \
        neo4j:latest
    ```

    Args:
        username (str): The username for the Neo4j database.
        password (str): The password for the Neo4j database.
        url (str): The URL for the Neo4j database.
        database (Optional[str]): The name of the database to connect to. Defaults to "neo4j".

    Examples:
        `pip install llama-index-graph-stores-neo4j`

        ```python
        from llama_index.core.indices.property_graph import PropertyGraphIndex
        from llama_index.graph_stores.neo4j import Neo4jPropertyGraphStore

        # Create a Neo4jPropertyGraphStore instance
        graph_store = Neo4jPropertyGraphStore(
            username="neo4j",
            password="neo4j",
            url="bolt://localhost:7687",
            database="neo4j"
        )

        # create the index
        index = PropertyGraphIndex.from_documents(
            documents,
            property_graph_store=graph_store,
        )
        ```
    """

    supports_structured_queries: bool = True
    supports_vector_queries: bool = True
    text_to_cypher_template: PromptTemplate = DEFAULT_CYPHER_TEMPALTE

    def __init__(
        self,
        username: str,
        password: str,
        url: str,
        database: Optional[str] = "neo4j",
        refresh_schema: bool = True,
        sanitize_query_output: bool = True,
        enhanced_schema: bool = False,
        **neo4j_kwargs: Any,
    ) -> None:
        self.sanitize_query_output = sanitize_query_output
        self.enhanced_schema = enhanced_schema
        self._driver = neo4j.GraphDatabase.driver(
            url, auth=(username, password), **neo4j_kwargs
        )
        self._async_driver = neo4j.AsyncGraphDatabase.driver(
            url,
            auth=(username, password),
            **neo4j_kwargs,
        )
        self._database = database
        self.structured_schema = {}
        if refresh_schema:
            self.refresh_schema()

    @property
    def client(self):
        return self._driver

    def refresh_schema(self) -> None:
        """Refresh the schema."""
        node_query_results = self.structured_query(
            node_properties_query,
            param_map={"EXCLUDED_LABELS": [*EXCLUDED_LABELS, BASE_ENTITY_LABEL]},
        )
        node_properties = (
            [el["output"] for el in node_query_results] if node_query_results else []
        )

        rels_query_result = self.structured_query(
            rel_properties_query, param_map={"EXCLUDED_LABELS": EXCLUDED_RELS}
        )
        rel_properties = (
            [el["output"] for el in rels_query_result] if rels_query_result else []
        )

        rel_objs_query_result = self.structured_query(
            rel_query,
            param_map={"EXCLUDED_LABELS": [*EXCLUDED_LABELS, BASE_ENTITY_LABEL]},
        )
        relationships = (
            [el["output"] for el in rel_objs_query_result]
            if rel_objs_query_result
            else []
        )

        # Get constraints & indexes
        try:
            constraint = self.structured_query("SHOW CONSTRAINTS")
            index = self.structured_query(
                "CALL apoc.schema.nodes() YIELD label, properties, type, size, "
                "valuesSelectivity WHERE type = 'RANGE' RETURN *, "
                "size * valuesSelectivity as distinctValues"
            )
        except (
            neo4j.exceptions.ClientError
        ):  # Read-only user might not have access to schema information
            constraint = []
            index = []

        self.structured_schema = {
            "node_props": {el["labels"]: el["properties"] for el in node_properties},
            "rel_props": {el["type"]: el["properties"] for el in rel_properties},
            "relationships": relationships,
            "metadata": {"constraint": constraint, "index": index},
        }
        schema_counts = self.structured_query(
            "CALL apoc.meta.graphSample() YIELD nodes, relationships "
            "RETURN nodes, [rel in relationships | {name:apoc.any.property"
            "(rel, 'type'), count: apoc.any.property(rel, 'count')}]"
            " AS relationships"
        )
        # Update node info
        for node in schema_counts[0].get("nodes", []):
            # Skip bloom labels
            if node["name"] in EXCLUDED_LABELS:
                continue
            node_props = self.structured_schema["node_props"].get(node["name"])
            if not node_props:  # The node has no properties
                continue
            enhanced_cypher = self._enhanced_schema_cypher(
                node["name"], node_props, node["count"] < EXHAUSTIVE_SEARCH_LIMIT
            )
            enhanced_info = self.structured_query(enhanced_cypher)[0]["output"]
            for prop in node_props:
                if prop["property"] in enhanced_info:
                    prop.update(enhanced_info[prop["property"]])
        # Update rel info
        for rel in schema_counts[0].get("relationships", []):
            # Skip bloom labels
            if rel["name"] in EXCLUDED_RELS:
                continue
            rel_props = self.structured_schema["rel_props"].get(rel["name"])
            if not rel_props:  # The rel has no properties
                continue
            enhanced_cypher = self._enhanced_schema_cypher(
                rel["name"],
                rel_props,
                rel["count"] < EXHAUSTIVE_SEARCH_LIMIT,
                is_relationship=True,
            )
            try:
                enhanced_info = self.structured_query(enhanced_cypher)[0]["output"]
                for prop in rel_props:
                    if prop["property"] in enhanced_info:
                        prop.update(enhanced_info[prop["property"]])
            except neo4j.exceptions.ClientError:
                # Sometimes the types are not consistent in the db
                pass

    def upsert_nodes(self, nodes: List[LabelledNode]) -> None:
        # Lists to hold separated types
        entity_dicts: List[dict] = []
        chunk_dicts: List[dict] = []

        # Sort by type
        for item in nodes:
            if isinstance(item, EntityNode):
                entity_dicts.append({**item.dict(), "id": item.id})
            elif isinstance(item, ChunkNode):
                chunk_dicts.append({**item.dict(), "id": item.id})
            else:
                # Log that we do not support these types of nodes
                # Or raise an error?
                pass

        if chunk_dicts:
            self.structured_query(
                """
                UNWIND $data AS row
                MERGE (c:Chunk {id: row.id})
                SET c.text = row.text
                WITH c, row
                SET c += row.properties
                WITH c, row.embedding AS embedding
                WHERE embedding IS NOT NULL
                CALL db.create.setNodeVectorProperty(c, 'embedding', embedding)
                RETURN count(*)
                """,
                param_map={"data": chunk_dicts},
            )

        if entity_dicts:
            self.structured_query(
                """
                UNWIND $data AS row
                MERGE (e:`__Entity__` {id: row.id})
                SET e += apoc.map.clean(row.properties, [], [])
                SET e.name = row.name
                WITH e, row
                CALL apoc.create.addLabels(e, [row.label])
                YIELD node
                WITH e, row
                CALL {
                    WITH e, row
                    WITH e, row
                    WHERE row.embedding IS NOT NULL
                    CALL db.create.setNodeVectorProperty(e, 'embedding', row.embedding)
                    RETURN count(*) AS count
                }
                WITH e, row WHERE row.properties.triplet_source_id IS NOT NULL
                MERGE (c:Chunk {id: row.properties.triplet_source_id})
                MERGE (e)<-[:MENTIONS]-(c)
                """,
                param_map={"data": entity_dicts},
            )

    def upsert_relations(self, relations: List[Relation]) -> None:
        """Add relations."""
        params = [r.dict() for r in relations]

        self.structured_query(
            """
            UNWIND $data AS row
            MERGE (source {id: row.source_id})
            ON CREATE SET source:Chunk
            MERGE (target {id: row.target_id})
            ON CREATE SET target:Chunk
            WITH source, target, row
            CALL apoc.merge.relationship(source, row.label, {}, row.properties, target) YIELD rel
            RETURN count(*)
            """,
            param_map={"data": params},
        )

    def get(
        self,
        properties: Optional[dict] = None,
        ids: Optional[List[str]] = None,
    ) -> List[LabelledNode]:
        """Get nodes."""
        cypher_statement = "MATCH (e) "

        params = {}
        if properties or ids:
            cypher_statement += "WHERE "

        if ids:
            cypher_statement += "e.id in $ids "
            params["ids"] = ids

        if properties:
            prop_list = []
            for i, prop in enumerate(properties):
                prop_list.append(f"e.`{prop}` = $property_{i}")
                params[f"property_{i}"] = properties[prop]
            cypher_statement += " AND ".join(prop_list)

        return_statement = """
        WITH e
        RETURN e.id AS name,
               [l in labels(e) WHERE l <> '__Entity__' | l][0] AS type,
               e{.* , embedding: Null, id: Null} AS properties
        """
        cypher_statement += return_statement

        response = self.structured_query(cypher_statement, param_map=params)
        response = response if response else []

        nodes = []
        for record in response:
            # text indicates a chunk node
            # none on the type indicates an implicit node, likely a chunk node
            if "text" in record["properties"] or record["type"] is None:
                text = record["properties"].pop("text", "")
                nodes.append(
                    ChunkNode(
                        id_=record["name"],
                        text=text,
                        properties=remove_empty_values(record["properties"]),
                    )
                )
            else:
                nodes.append(
                    EntityNode(
                        name=record["name"],
                        label=record["type"],
                        properties=remove_empty_values(record["properties"]),
                    )
                )

        return nodes

    def get_triplets(
        self,
        entity_names: Optional[List[str]] = None,
        relation_names: Optional[List[str]] = None,
        properties: Optional[dict] = None,
        ids: Optional[List[str]] = None,
    ) -> List[Triplet]:
        # TODO: handle ids of chunk nodes
        cypher_statement = "MATCH (e:`__Entity__`) "

        params = {}
        if entity_names or properties or ids:
            cypher_statement += "WHERE "

        if entity_names:
            cypher_statement += "e.name in $entity_names "
            params["entity_names"] = entity_names

        if ids:
            cypher_statement += "e.id in $ids "
            params["ids"] = ids

        if properties:
            prop_list = []
            for i, prop in enumerate(properties):
                prop_list.append(f"e.`{prop}` = $property_{i}")
                params[f"property_{i}"] = properties[prop]
            cypher_statement += " AND ".join(prop_list)

        return_statement = f"""
        WITH e
        CALL {{
            WITH e
            MATCH (e)-[r{':`' + '`|`'.join(relation_names) + '`' if relation_names else ''}]->(t)
            RETURN e.name AS source_id, [l in labels(e) WHERE l <> '__Entity__' | l][0] AS source_type,
                   e{{.* , embedding: Null, name: Null}} AS source_properties,
                   type(r) AS type,
                   t.name AS target_id, [l in labels(t) WHERE l <> '__Entity__' | l][0] AS target_type,
                   t{{.* , embedding: Null, name: Null}} AS target_properties
            UNION ALL
            WITH e
            MATCH (e)<-[r{':`' + '`|`'.join(relation_names) + '`' if relation_names else ''}]-(t)
            RETURN t.name AS source_id, [l in labels(t) WHERE l <> '__Entity__' | l][0] AS source_type,
                   e{{.* , embedding: Null, name: Null}} AS source_properties,
                   type(r) AS type,
                   e.name AS target_id, [l in labels(e) WHERE l <> '__Entity__' | l][0] AS target_type,
                   t{{.* , embedding: Null, name: Null}} AS target_properties
        }}
        RETURN source_id, source_type, type, target_id, target_type, source_properties, target_properties"""
        cypher_statement += return_statement

        data = self.structured_query(cypher_statement, param_map=params)
        data = data if data else []

        triples = []
        for record in data:
            source = EntityNode(
                name=record["source_id"],
                label=record["source_type"],
                properties=remove_empty_values(record["source_properties"]),
            )
            target = EntityNode(
                name=record["target_id"],
                label=record["target_type"],
                properties=remove_empty_values(record["target_properties"]),
            )
            rel = Relation(
                source_id=record["source_id"],
                target_id=record["target_id"],
                label=record["type"],
            )
            triples.append([source, rel, target])
        return triples

    def get_rel_map(
        self,
        graph_nodes: List[LabelledNode],
        depth: int = 2,
        limit: int = 30,
        ignore_rels: Optional[List[str]] = None,
    ) -> List[Triplet]:
        """Get depth-aware rel map."""
        triples = []

        ids = [node.id for node in graph_nodes]
        # Needs some optimization
        response = self.structured_query(
            f"""
            WITH $ids AS id_list
            UNWIND range(0, size(id_list) - 1) AS idx
            MATCH (e:`__Entity__`)
            WHERE e.id = id_list[idx]
            MATCH p=(e)-[r*1..{depth}]-(other)
            WHERE ALL(rel in relationships(p) WHERE type(rel) <> 'MENTIONS')
            UNWIND relationships(p) AS rel
            WITH distinct rel, idx
            WITH startNode(rel) AS source,
                type(rel) AS type,
                endNode(rel) AS endNode,
                idx
            LIMIT toInteger($limit)
            RETURN source.id AS source_id, [l in labels(source) WHERE l <> '__Entity__' | l][0] AS source_type,
                source{{.* , embedding: Null, id: Null}} AS source_properties,
                type,
                endNode.id AS target_id, [l in labels(endNode) WHERE l <> '__Entity__' | l][0] AS target_type,
                endNode{{.* , embedding: Null, id: Null}} AS target_properties,
                idx
            ORDER BY idx
            LIMIT toInteger($limit)
            """,
            param_map={"ids": ids, "limit": limit},
        )
        response = response if response else []

        ignore_rels = ignore_rels or []
        for record in response:
            if record["type"] in ignore_rels:
                continue

            source = EntityNode(
                name=record["source_id"],
                label=record["source_type"],
                properties=remove_empty_values(record["source_properties"]),
            )
            target = EntityNode(
                name=record["target_id"],
                label=record["target_type"],
                properties=remove_empty_values(record["target_properties"]),
            )
            rel = Relation(
                source_id=record["source_id"],
                target_id=record["target_id"],
                label=record["type"],
            )
            triples.append([source, rel, target])

        return triples

    def structured_query(
        self, query: str, param_map: Optional[Dict[str, Any]] = None
    ) -> Any:
        param_map = param_map or {}

        with self._driver.session(database=self._database) as session:
            result = session.run(query, param_map)
            full_result = [d.data() for d in result]

        if self.sanitize_query_output:
            return value_sanitize(full_result)

        return full_result

    def vector_query(
        self, query: VectorStoreQuery, **kwargs: Any
    ) -> Tuple[List[LabelledNode], List[float]]:
        """Query the graph store with a vector store query."""
        conditions = None
        if query.filters:
            conditions = [
                f"e.{filter.key} {filter.operator.value} {filter.value}"
                for filter in query.filters.filters
            ]
        filters = (
            f" {query.filters.condition.value} ".join(conditions).replace("==", "=")
            if conditions is not None
            else "1 = 1"
        )

        data = self.structured_query(
            f"""MATCH (e:`__Entity__`)
            WHERE e.embedding IS NOT NULL AND size(e.embedding) = $dimension AND ({filters})
            WITH e, vector.similarity.cosine(e.embedding, $embedding) AS score
            ORDER BY score DESC LIMIT toInteger($limit)
            RETURN e.id AS name,
               [l in labels(e) WHERE l <> '__Entity__' | l][0] AS type,
               e{{.* , embedding: Null, name: Null, id: Null}} AS properties,
               score""",
            param_map={
                "embedding": query.query_embedding,
                "dimension": len(query.query_embedding),
                "limit": query.similarity_top_k,
            },
        )
        data = data if data else []

        nodes = []
        scores = []
        for record in data:
            node = EntityNode(
                name=record["name"],
                label=record["type"],
                properties=remove_empty_values(record["properties"]),
            )
            nodes.append(node)
            scores.append(record["score"])

        return (nodes, scores)

    def delete(
        self,
        entity_names: Optional[List[str]] = None,
        relation_names: Optional[List[str]] = None,
        properties: Optional[dict] = None,
        ids: Optional[List[str]] = None,
    ) -> None:
        """Delete matching data."""
        if entity_names:
            self.structured_query(
                "MATCH (n) WHERE n.name IN $entity_names DETACH DELETE n",
                param_map={"entity_names": entity_names},
            )

        if ids:
            self.structured_query(
                "MATCH (n) WHERE n.id IN $ids DETACH DELETE n",
                param_map={"ids": ids},
            )

        if relation_names:
            for rel in relation_names:
                self.structured_query(f"MATCH ()-[r:`{rel}`]->() DELETE r")

        if properties:
            cypher = "MATCH (e) WHERE "
            prop_list = []
            params = {}
            for i, prop in enumerate(properties):
                prop_list.append(f"e.`{prop}` = $property_{i}")
                params[f"property_{i}"] = properties[prop]
            cypher += " AND ".join(prop_list)
            self.structured_query(cypher + " DETACH DELETE e", param_map=params)

    def _enhanced_schema_cypher(
        self,
        label_or_type: str,
        properties: List[Dict[str, Any]],
        exhaustive: bool,
        is_relationship: bool = False,
    ) -> str:
        if is_relationship:
            match_clause = f"MATCH ()-[n:`{label_or_type}`]->()"
        else:
            match_clause = f"MATCH (n:`{label_or_type}`)"

        with_clauses = []
        return_clauses = []
        output_dict = {}
        if exhaustive:
            for prop in properties:
                prop_name = prop["property"]
                prop_type = prop["type"]
                if prop_type == "STRING":
                    with_clauses.append(
                        f"collect(distinct substring(toString(n.`{prop_name}`), 0, 50)) "
                        f"AS `{prop_name}_values`"
                    )
                    return_clauses.append(
                        f"values:`{prop_name}_values`[..{DISTINCT_VALUE_LIMIT}],"
                        f" distinct_count: size(`{prop_name}_values`)"
                    )
                elif prop_type in [
                    "INTEGER",
                    "FLOAT",
                    "DATE",
                    "DATE_TIME",
                    "LOCAL_DATE_TIME",
                ]:
                    with_clauses.append(f"min(n.`{prop_name}`) AS `{prop_name}_min`")
                    with_clauses.append(f"max(n.`{prop_name}`) AS `{prop_name}_max`")
                    with_clauses.append(
                        f"count(distinct n.`{prop_name}`) AS `{prop_name}_distinct`"
                    )
                    return_clauses.append(
                        f"min: toString(`{prop_name}_min`), "
                        f"max: toString(`{prop_name}_max`), "
                        f"distinct_count: `{prop_name}_distinct`"
                    )
                elif prop_type == "LIST":
                    with_clauses.append(
                        f"min(size(n.`{prop_name}`)) AS `{prop_name}_size_min`, "
                        f"max(size(n.`{prop_name}`)) AS `{prop_name}_size_max`"
                    )
                    return_clauses.append(
                        f"min_size: `{prop_name}_size_min`, "
                        f"max_size: `{prop_name}_size_max`"
                    )
                elif prop_type in ["BOOLEAN", "POINT", "DURATION"]:
                    continue
                output_dict[prop_name] = "{" + return_clauses.pop() + "}"
        else:
            # Just sample 5 random nodes
            match_clause += " WITH n LIMIT 5"
            for prop in properties:
                prop_name = prop["property"]
                prop_type = prop["type"]

                # Check if indexed property, we can still do exhaustive
                prop_index = [
                    el
                    for el in self.structured_schema["metadata"]["index"]
                    if el["label"] == label_or_type
                    and el["properties"] == [prop_name]
                    and el["type"] == "RANGE"
                ]
                if prop_type == "STRING":
                    if (
                        prop_index
                        and prop_index[0].get("size") > 0
                        and prop_index[0].get("distinctValues") <= DISTINCT_VALUE_LIMIT
                    ):
                        distinct_values = self.query(
                            f"CALL apoc.schema.properties.distinct("
                            f"'{label_or_type}', '{prop_name}') YIELD value"
                        )[0]["value"]
                        return_clauses.append(
                            f"values: {distinct_values},"
                            f" distinct_count: {len(distinct_values)}"
                        )
                    else:
                        with_clauses.append(
                            f"collect(distinct substring(n.`{prop_name}`, 0, 50)) "
                            f"AS `{prop_name}_values`"
                        )
                        return_clauses.append(f"values: `{prop_name}_values`")
                elif prop_type in [
                    "INTEGER",
                    "FLOAT",
                    "DATE",
                    "DATE_TIME",
                    "LOCAL_DATE_TIME",
                ]:
                    if not prop_index:
                        with_clauses.append(
                            f"collect(distinct toString(n.`{prop_name}`)) "
                            f"AS `{prop_name}_values`"
                        )
                        return_clauses.append(f"values: `{prop_name}_values`")
                    else:
                        with_clauses.append(
                            f"min(n.`{prop_name}`) AS `{prop_name}_min`"
                        )
                        with_clauses.append(
                            f"max(n.`{prop_name}`) AS `{prop_name}_max`"
                        )
                        with_clauses.append(
                            f"count(distinct n.`{prop_name}`) AS `{prop_name}_distinct`"
                        )
                        return_clauses.append(
                            f"min: toString(`{prop_name}_min`), "
                            f"max: toString(`{prop_name}_max`), "
                            f"distinct_count: `{prop_name}_distinct`"
                        )

                elif prop_type == "LIST":
                    with_clauses.append(
                        f"min(size(n.`{prop_name}`)) AS `{prop_name}_size_min`, "
                        f"max(size(n.`{prop_name}`)) AS `{prop_name}_size_max`"
                    )
                    return_clauses.append(
                        f"min_size: `{prop_name}_size_min`, "
                        f"max_size: `{prop_name}_size_max`"
                    )
                elif prop_type in ["BOOLEAN", "POINT", "DURATION"]:
                    continue

                output_dict[prop_name] = "{" + return_clauses.pop() + "}"

        with_clause = "WITH " + ",\n     ".join(with_clauses)
        return_clause = (
            "RETURN {"
            + ", ".join(f"`{k}`: {v}" for k, v in output_dict.items())
            + "} AS output"
        )

        # Combine all parts of the Cypher query
        return f"{match_clause}\n{with_clause}\n{return_clause}"

    def get_schema(self, refresh: bool = False) -> Any:
        if refresh:
            self.refresh_schema()

        return self.structured_schema

    def get_schema_str(self, refresh: bool = False) -> str:
        schema = self.get_schema(refresh=refresh)

        formatted_node_props = []
        formatted_rel_props = []

        if self.enhanced_schema:
            # Enhanced formatting for nodes
            for node_type, properties in schema["node_props"].items():
                formatted_node_props.append(f"- **{node_type}**")
                for prop in properties:
                    example = ""
                    if prop["type"] == "STRING" and prop.get("values"):
                        if prop.get("distinct_count", 11) > DISTINCT_VALUE_LIMIT:
                            example = (
                                f'Example: "{clean_string_values(prop["values"][0])}"'
                                if prop["values"]
                                else ""
                            )
                        else:  # If less than 10 possible values return all
                            example = (
                                (
                                    "Available options: "
                                    f'{[clean_string_values(el) for el in prop["values"]]}'
                                )
                                if prop["values"]
                                else ""
                            )

                    elif prop["type"] in [
                        "INTEGER",
                        "FLOAT",
                        "DATE",
                        "DATE_TIME",
                        "LOCAL_DATE_TIME",
                    ]:
                        if prop.get("min") is not None:
                            example = f'Min: {prop["min"]}, Max: {prop["max"]}'
                        else:
                            example = (
                                f'Example: "{prop["values"][0]}"'
                                if prop.get("values")
                                else ""
                            )
                    elif prop["type"] == "LIST":
                        # Skip embeddings
                        if not prop.get("min_size") or prop["min_size"] > LIST_LIMIT:
                            continue
                        example = f'Min Size: {prop["min_size"]}, Max Size: {prop["max_size"]}'
                    formatted_node_props.append(
                        f"  - `{prop['property']}`: {prop['type']} {example}"
                    )

            # Enhanced formatting for relationships
            for rel_type, properties in schema["rel_props"].items():
                formatted_rel_props.append(f"- **{rel_type}**")
                for prop in properties:
                    example = ""
                    if prop["type"] == "STRING":
                        if prop.get("distinct_count", 11) > DISTINCT_VALUE_LIMIT:
                            example = (
                                f'Example: "{clean_string_values(prop["values"][0])}"'
                                if prop.get("values")
                                else ""
                            )
                        else:  # If less than 10 possible values return all
                            example = (
                                (
                                    "Available options: "
                                    f'{[clean_string_values(el) for el in prop["values"]]}'
                                )
                                if prop.get("values")
                                else ""
                            )
                    elif prop["type"] in [
                        "INTEGER",
                        "FLOAT",
                        "DATE",
                        "DATE_TIME",
                        "LOCAL_DATE_TIME",
                    ]:
                        if prop.get("min"):  # If we have min/max
                            example = f'Min: {prop["min"]}, Max:  {prop["max"]}'
                        else:  # return a single value
                            example = (
                                f'Example: "{prop["values"][0]}"'
                                if prop.get("values")
                                else ""
                            )
                    elif prop["type"] == "LIST":
                        # Skip embeddings
                        if prop["min_size"] > LIST_LIMIT:
                            continue
                        example = f'Min Size: {prop["min_size"]}, Max Size: {prop["max_size"]}'
                    formatted_rel_props.append(
                        f"  - `{prop['property']}: {prop['type']}` {example}"
                    )
        else:
            # Format node properties
            for label, props in schema["node_props"].items():
                props_str = ", ".join(
                    [f"{prop['property']}: {prop['type']}" for prop in props]
                )
                formatted_node_props.append(f"{label} {{{props_str}}}")

            # Format relationship properties using structured_schema
            for type, props in schema["rel_props"].items():
                props_str = ", ".join(
                    [f"{prop['property']}: {prop['type']}" for prop in props]
                )
                formatted_rel_props.append(f"{type} {{{props_str}}}")

        # Format relationships
        formatted_rels = [
            f"(:{el['start']})-[:{el['type']}]->(:{el['end']})"
            for el in schema["relationships"]
        ]

        return "\n".join(
            [
                "Node properties:",
                "\n".join(formatted_node_props),
                "Relationship properties:",
                "\n".join(formatted_rel_props),
                "The relationships:",
                "\n".join(formatted_rels),
            ]
        )

refresh_schema #

refresh_schema() -> None

Refresh the schema.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
def refresh_schema(self) -> None:
    """Refresh the schema."""
    node_query_results = self.structured_query(
        node_properties_query,
        param_map={"EXCLUDED_LABELS": [*EXCLUDED_LABELS, BASE_ENTITY_LABEL]},
    )
    node_properties = (
        [el["output"] for el in node_query_results] if node_query_results else []
    )

    rels_query_result = self.structured_query(
        rel_properties_query, param_map={"EXCLUDED_LABELS": EXCLUDED_RELS}
    )
    rel_properties = (
        [el["output"] for el in rels_query_result] if rels_query_result else []
    )

    rel_objs_query_result = self.structured_query(
        rel_query,
        param_map={"EXCLUDED_LABELS": [*EXCLUDED_LABELS, BASE_ENTITY_LABEL]},
    )
    relationships = (
        [el["output"] for el in rel_objs_query_result]
        if rel_objs_query_result
        else []
    )

    # Get constraints & indexes
    try:
        constraint = self.structured_query("SHOW CONSTRAINTS")
        index = self.structured_query(
            "CALL apoc.schema.nodes() YIELD label, properties, type, size, "
            "valuesSelectivity WHERE type = 'RANGE' RETURN *, "
            "size * valuesSelectivity as distinctValues"
        )
    except (
        neo4j.exceptions.ClientError
    ):  # Read-only user might not have access to schema information
        constraint = []
        index = []

    self.structured_schema = {
        "node_props": {el["labels"]: el["properties"] for el in node_properties},
        "rel_props": {el["type"]: el["properties"] for el in rel_properties},
        "relationships": relationships,
        "metadata": {"constraint": constraint, "index": index},
    }
    schema_counts = self.structured_query(
        "CALL apoc.meta.graphSample() YIELD nodes, relationships "
        "RETURN nodes, [rel in relationships | {name:apoc.any.property"
        "(rel, 'type'), count: apoc.any.property(rel, 'count')}]"
        " AS relationships"
    )
    # Update node info
    for node in schema_counts[0].get("nodes", []):
        # Skip bloom labels
        if node["name"] in EXCLUDED_LABELS:
            continue
        node_props = self.structured_schema["node_props"].get(node["name"])
        if not node_props:  # The node has no properties
            continue
        enhanced_cypher = self._enhanced_schema_cypher(
            node["name"], node_props, node["count"] < EXHAUSTIVE_SEARCH_LIMIT
        )
        enhanced_info = self.structured_query(enhanced_cypher)[0]["output"]
        for prop in node_props:
            if prop["property"] in enhanced_info:
                prop.update(enhanced_info[prop["property"]])
    # Update rel info
    for rel in schema_counts[0].get("relationships", []):
        # Skip bloom labels
        if rel["name"] in EXCLUDED_RELS:
            continue
        rel_props = self.structured_schema["rel_props"].get(rel["name"])
        if not rel_props:  # The rel has no properties
            continue
        enhanced_cypher = self._enhanced_schema_cypher(
            rel["name"],
            rel_props,
            rel["count"] < EXHAUSTIVE_SEARCH_LIMIT,
            is_relationship=True,
        )
        try:
            enhanced_info = self.structured_query(enhanced_cypher)[0]["output"]
            for prop in rel_props:
                if prop["property"] in enhanced_info:
                    prop.update(enhanced_info[prop["property"]])
        except neo4j.exceptions.ClientError:
            # Sometimes the types are not consistent in the db
            pass

upsert_relations #

upsert_relations(relations: List[Relation]) -> None

Add relations.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
def upsert_relations(self, relations: List[Relation]) -> None:
    """Add relations."""
    params = [r.dict() for r in relations]

    self.structured_query(
        """
        UNWIND $data AS row
        MERGE (source {id: row.source_id})
        ON CREATE SET source:Chunk
        MERGE (target {id: row.target_id})
        ON CREATE SET target:Chunk
        WITH source, target, row
        CALL apoc.merge.relationship(source, row.label, {}, row.properties, target) YIELD rel
        RETURN count(*)
        """,
        param_map={"data": params},
    )

get #

get(properties: Optional[dict] = None, ids: Optional[List[str]] = None) -> List[LabelledNode]

Get nodes.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
def get(
    self,
    properties: Optional[dict] = None,
    ids: Optional[List[str]] = None,
) -> List[LabelledNode]:
    """Get nodes."""
    cypher_statement = "MATCH (e) "

    params = {}
    if properties or ids:
        cypher_statement += "WHERE "

    if ids:
        cypher_statement += "e.id in $ids "
        params["ids"] = ids

    if properties:
        prop_list = []
        for i, prop in enumerate(properties):
            prop_list.append(f"e.`{prop}` = $property_{i}")
            params[f"property_{i}"] = properties[prop]
        cypher_statement += " AND ".join(prop_list)

    return_statement = """
    WITH e
    RETURN e.id AS name,
           [l in labels(e) WHERE l <> '__Entity__' | l][0] AS type,
           e{.* , embedding: Null, id: Null} AS properties
    """
    cypher_statement += return_statement

    response = self.structured_query(cypher_statement, param_map=params)
    response = response if response else []

    nodes = []
    for record in response:
        # text indicates a chunk node
        # none on the type indicates an implicit node, likely a chunk node
        if "text" in record["properties"] or record["type"] is None:
            text = record["properties"].pop("text", "")
            nodes.append(
                ChunkNode(
                    id_=record["name"],
                    text=text,
                    properties=remove_empty_values(record["properties"]),
                )
            )
        else:
            nodes.append(
                EntityNode(
                    name=record["name"],
                    label=record["type"],
                    properties=remove_empty_values(record["properties"]),
                )
            )

    return nodes

get_rel_map #

get_rel_map(graph_nodes: List[LabelledNode], depth: int = 2, limit: int = 30, ignore_rels: Optional[List[str]] = None) -> List[Triplet]

Get depth-aware rel map.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
def get_rel_map(
    self,
    graph_nodes: List[LabelledNode],
    depth: int = 2,
    limit: int = 30,
    ignore_rels: Optional[List[str]] = None,
) -> List[Triplet]:
    """Get depth-aware rel map."""
    triples = []

    ids = [node.id for node in graph_nodes]
    # Needs some optimization
    response = self.structured_query(
        f"""
        WITH $ids AS id_list
        UNWIND range(0, size(id_list) - 1) AS idx
        MATCH (e:`__Entity__`)
        WHERE e.id = id_list[idx]
        MATCH p=(e)-[r*1..{depth}]-(other)
        WHERE ALL(rel in relationships(p) WHERE type(rel) <> 'MENTIONS')
        UNWIND relationships(p) AS rel
        WITH distinct rel, idx
        WITH startNode(rel) AS source,
            type(rel) AS type,
            endNode(rel) AS endNode,
            idx
        LIMIT toInteger($limit)
        RETURN source.id AS source_id, [l in labels(source) WHERE l <> '__Entity__' | l][0] AS source_type,
            source{{.* , embedding: Null, id: Null}} AS source_properties,
            type,
            endNode.id AS target_id, [l in labels(endNode) WHERE l <> '__Entity__' | l][0] AS target_type,
            endNode{{.* , embedding: Null, id: Null}} AS target_properties,
            idx
        ORDER BY idx
        LIMIT toInteger($limit)
        """,
        param_map={"ids": ids, "limit": limit},
    )
    response = response if response else []

    ignore_rels = ignore_rels or []
    for record in response:
        if record["type"] in ignore_rels:
            continue

        source = EntityNode(
            name=record["source_id"],
            label=record["source_type"],
            properties=remove_empty_values(record["source_properties"]),
        )
        target = EntityNode(
            name=record["target_id"],
            label=record["target_type"],
            properties=remove_empty_values(record["target_properties"]),
        )
        rel = Relation(
            source_id=record["source_id"],
            target_id=record["target_id"],
            label=record["type"],
        )
        triples.append([source, rel, target])

    return triples

vector_query #

vector_query(query: VectorStoreQuery, **kwargs: Any) -> Tuple[List[LabelledNode], List[float]]

Query the graph store with a vector store query.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
def vector_query(
    self, query: VectorStoreQuery, **kwargs: Any
) -> Tuple[List[LabelledNode], List[float]]:
    """Query the graph store with a vector store query."""
    conditions = None
    if query.filters:
        conditions = [
            f"e.{filter.key} {filter.operator.value} {filter.value}"
            for filter in query.filters.filters
        ]
    filters = (
        f" {query.filters.condition.value} ".join(conditions).replace("==", "=")
        if conditions is not None
        else "1 = 1"
    )

    data = self.structured_query(
        f"""MATCH (e:`__Entity__`)
        WHERE e.embedding IS NOT NULL AND size(e.embedding) = $dimension AND ({filters})
        WITH e, vector.similarity.cosine(e.embedding, $embedding) AS score
        ORDER BY score DESC LIMIT toInteger($limit)
        RETURN e.id AS name,
           [l in labels(e) WHERE l <> '__Entity__' | l][0] AS type,
           e{{.* , embedding: Null, name: Null, id: Null}} AS properties,
           score""",
        param_map={
            "embedding": query.query_embedding,
            "dimension": len(query.query_embedding),
            "limit": query.similarity_top_k,
        },
    )
    data = data if data else []

    nodes = []
    scores = []
    for record in data:
        node = EntityNode(
            name=record["name"],
            label=record["type"],
            properties=remove_empty_values(record["properties"]),
        )
        nodes.append(node)
        scores.append(record["score"])

    return (nodes, scores)

delete #

delete(entity_names: Optional[List[str]] = None, relation_names: Optional[List[str]] = None, properties: Optional[dict] = None, ids: Optional[List[str]] = None) -> None

Delete matching data.

Source code in llama-index-integrations/graph_stores/llama-index-graph-stores-neo4j/llama_index/graph_stores/neo4j/neo4j_property_graph.py
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
def delete(
    self,
    entity_names: Optional[List[str]] = None,
    relation_names: Optional[List[str]] = None,
    properties: Optional[dict] = None,
    ids: Optional[List[str]] = None,
) -> None:
    """Delete matching data."""
    if entity_names:
        self.structured_query(
            "MATCH (n) WHERE n.name IN $entity_names DETACH DELETE n",
            param_map={"entity_names": entity_names},
        )

    if ids:
        self.structured_query(
            "MATCH (n) WHERE n.id IN $ids DETACH DELETE n",
            param_map={"ids": ids},
        )

    if relation_names:
        for rel in relation_names:
            self.structured_query(f"MATCH ()-[r:`{rel}`]->() DELETE r")

    if properties:
        cypher = "MATCH (e) WHERE "
        prop_list = []
        params = {}
        for i, prop in enumerate(properties):
            prop_list.append(f"e.`{prop}` = $property_{i}")
            params[f"property_{i}"] = properties[prop]
        cypher += " AND ".join(prop_list)
        self.structured_query(cypher + " DETACH DELETE e", param_map=params)